Gemma

Gemma is Google’s family of open models built from the same research and technology behind the Gemini models. Gemma 4 features advanced reasoning with multi-step planning, native multimodal understanding (video, image, and audio), agentic capabilities with function-calling and structured JSON output, and 256K context windows — all under the Apache 2.0 license. Models range from edge-optimized E2B/E4B for mobile and IoT to 26B MoE and 31B dense variants for research and production workloads.